Jev with LiteLLM

LiteLLM lets your apps talk to one proxy instead of many provider APIs. Jev is a different kind of call — a typed decision ({ state, questions } in, a typed answers map out, not a chat completion) — so you route it through LiteLLM as a pass-through. Point a route at the jevtypesafeai.com gateway with your jv_live_ key and Jev sits behind the same base URL as your chat models, billed through your hosted account.

A Jev call isn't an OpenAI-style chat completion. You POST { state, questions } and get back a typed answers map — one of your options, a calibrated probability for each, and a confidence value, in roughly 70–500ms. LiteLLM handles that cleanly as a pass-through endpoint: the proxy forwards your request body straight to Jev and returns the typed response, so your app keeps a single base URL for everything. Don't try to map a Jev call onto the chat-completions schema — route it as-is.

Route through your hosted Jev account (bills you)

Add a pass-through route to your LiteLLM proxy config that targets the hosted gateway, with your jv_live_ key supplied by the proxy:

# litellm config.yaml — expose /jev and forward it to the hosted gateway
general_settings:
  pass_through_endpoints:
    - path: "/jev"
      target: "https://jevtypesafeai.com/api/v1/decide"
      headers:
        Authorization: "Bearer os.environ/JEV_API_KEY"   # jv_live_...
        content-type: "application/json"

Now your apps POST the normal Jev body to the proxy and read the typed answer back — one base URL, billed through your hosted account:

const res = await fetch(`${LITELLM_PROXY_BASE_URL}/jev`, {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    state: { ticket: text },
    questions: {
      queue: {
        type: "choice",
        instructions: "Which queue should this go to?",
        criteria: { billing: "payments", technical: "bugs", abuse: "spam/fraud", sales: "pricing" },
      },
    },
  }),
});
const { answers } = await res.json();
route(answers.queue.choice); // same typed shape, now through your proxy

The built-in TypeSafe provider

LiteLLM also ships a TypeSafe (Jev) pass-through provider: you replace https://api.typesafe.ai with LITELLM_PROXY_BASE_URL/typesafe and set TYPESAFE_API_KEY in the proxy. Note this one targets the official api.typesafe.ai endpoint (with an official TypeSafe key), not your hosted jevtypesafeai.com account — so use the custom /jev route above when you want decisions billed through your hosted jv_live_ key.

Why front Jev with LiteLLM

• One base URL for your apps — Jev sits next to your chat models behind the same proxy.
• Central keys, budgets, rate limits and logging across every provider, Jev included.
• Swap or add decision backends without touching app code — just change the route target.
• The typed response is unchanged: you still read answers.<question>.choice / .noul / .score.

Billing is the same wherever you route it: with Jev, output is free and input is billed per million tokens on a sliding scale that gets cheaper the more you prepay — in practice a single decision costs well under a cent. Current tiers are on the pricing page.

FAQ

See also: How to use the Jev API · Jev and OpenRouter · Jev and Bifrost · Get a Jev key · Pricing